Thread (21 messages) 21 messages, 3 authors, 2015-07-08
DORMANTno replies

[PATCH 2/9] locking/qrwlock: avoid redundant atomic_add_return on read_lock_slowpath

From: Will Deacon <hidden>
Date: 2015-07-08 13:37:04
Also in: linux-arch

On Wed, Jul 08, 2015 at 10:59:44AM +0100, Peter Zijlstra wrote:
On Tue, Jul 07, 2015 at 03:28:13PM -0400, Waiman Long wrote:
quoted
#ifndef queued_inc_reader_return
static inline u32 queued_inc_reader_return(struct qrwlock *lock)
That's actually a misleading name, we don't do an inc_return, we do a
fetch_inc.
quoted
{
    return atomic_add_return(_QR_BIAS, &lock->cnts) - _QR_BIAS;
}
#endif
With that you (Will) could even reuse the ll/sc load and avoid issuing a
second load. You basically need atomic_fetch_add_relaxed().
ARMv8.1 has an instruction for that :) [ldadd].

Will
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help